How to Ascertain the Horizontal Pages - Assignment Example

Instrumentation Laboratory Report Number Question To ascertain the horizontal pages, the Prewitt_H filter detected on the grey level pictures horizontal directions and lines. In these cases, the images had whitish edges and black backgrounds. The lowest values (negatives) in the Prewitt_H filters are indicated with black colour whereas the maximum positive values are indicated in white colour. Along the same vein, in such a scenario, the black also denotes the zero value and in essence, black edges/ negative values (lines) are not discernible/ exhibited due to the background that is black. They show the same intensity with some degree of noise to them. Also, to ascertain the vertical edges, the Prewitt_V filter detected on the grey level images the vertical directions and lines with the backgrounds taking on black colours while the edges are white in these image types. In this case, the maximum positive values are white while the negative values are denoted in white. The zero value is black. In the end, the negative values are not seen/ discernible as they show equal intensity with the background. The Prewitt filters are rooted on the idea of central difference. As highlighted in figures 1c, 2c and 3c, the Prewitt filter pairs were applied to three (3) kinds of images with the resultant backgrounds of the set up filtered images transforming into grey colour. The Prewitt_H filter delves on the edges and lines that have horizontal directions and the circle image background inside the circle, that was previously white, is transformed to grey colour. In addition, the horizontal edges in the circle image have been enhanced and edges discerned/ detected. In the circle image, the top edges were of the lowest/ negative values, the black and bottom edges being of maximum/ positive values and the grey edges being the zero values with the filter response in areas of flat images being zero. The sharpness of the edges increases inside the circle and the circle is indented inside the image with some noise generally observed. The similar observation applied to the other two images (2c and 3c), in that the top edges were negative values/ black, the bottom edges positive values/ white and the zero edges remaining grey. Both the two above images had grey backgrounds too. The Prewitt_V filters target lines/ edges with vertical directions and it can be seen that the 3 images (1c, 2c and 3c) have been filtered using the Prewitt_V with the vertical lines in these scenarios greatly enhanced. In the circle image, that is figure 1c, the right side appears darker with the left side lighter. The Prewitt_V filter effectively increases the performance of the circle in the center of the square as the circle emerges on the peak of the square all because of the heightened sharpness. The positive values are lighter, the negative values lighter and the zero values exhibiting grey colour. The same results are seen in the remaining two images (2c and 3c) with all the backgrounds equally grey. The Sobel filter is composed of two kernels which detect horizontal and vertical variations/ changes in an image thus stressing on areas of high spatial frequency that correspond to the edges. If both kernels are used on an image, the results can be used to compute the magnitude and direction of the image edges. The images in figures 4c, 5c and 6c were created by applying Sobel_V and Sobel_H with similar results to the Prewitt filters. The Sobel filters can provide greater weight to the central pixels during averaging and in comparison to the Prewitt_H, the Sobel_H reduces the noise more. As we already understand, noise affects edge detection by giving erroneous edge positions and false edges altogether. The Roberts filters have small kernels which are derivatives in respect to the two diagonal directions. The Roberts filters are very sensitive to noise and cannot detect edges in the presence of noise therefore. These filters were used on a group of images as shown in figures 7c, 8c and 9c. In as far as all images exhibited sharper edges hence, some images were brighter than the others with other considered darker. These filters detect regions of pixel value gradient and the gradient can be found in different directions. The maximum positive values in such instances are white, the negative values black and the zero values grey. All these filters are anisotropic filters. The gradient is a vector with known magnitude and direction. The magnitude of the gradient is the amount of difference between the pixels in the neighborhood and the strength of the edge while the orientation/ direction of gradient provides the direction of greatest change which is basically the edge normal/ direction across the edge. Question 2 The above images (the Circle/ bmp, Yallingup/ jpg and Noise/ jpg) were then truncated using the Prewitt filters, that is, the Prewitt_H and the Prewitt_V. In these cases, the images displayed black backgrounds and edges that appeared white in colour. The images that are filtered with the Prewitt_H that have black colour have negative values whereas those in white are positive valued. In this case, the zero value is denoted by black and therefore the black edges/ lines are not indicated due to the black background. However, there is some noise in this situation. On the other hand, the Prewitt_V indicated similar results to the Prewitt_H in the sense that negative values were not shown/ visible due to similar intensities to the general background. This is simply to say that the Prewitt_H filter detected on the grey level images horizontal directions and lines to ascertain the horizontal edges whereas the Prewitt_V filter served to detect the vertical lines and directions on the grey level images to ascertain the vertical edges. Question 3 The Prewitt_H filter fails to detect the vertical edge when placed over this edge. This is so because it only detects the horizontal edge in the horizontal direction. Question 4 The Laplacian is a 2-D isotropic measure of the second spatial derivatives of an image. The Laplacian of an image indicates the areas of rapid change of intensity and is therefore often used for edge detection but is however very sensitive to noise. The Laplacian, in order to reduce its sensitivity to noise, is always used after image smoothing using a Gussian smoothing filter. The images in figure 1d and 2d are obtained following the application of the Laplacian filter to the bits (bmp) and the noisy image. The bits image (circle image) from the Laplacian filter has a lighter line that keeps the circle sharper and the amplifying of the noise does not greatly hamper the bits filtered image. The image details in the noisy image are lost with both images displaying grey backgrounds. The differences in the Laplacian image edges are due to the fact of constant intensity of the images. For example, when the intensity of a zero intensity gradient is altered, it responds in two ways – the positive values are relocated to the dark side whereas the negative values are located on the bright side. This is unlike other filters like the Sobel, Prewitt and the Roberts where positive values are white and the negative values black. Also, another notable difference is that the Laplacian filters can amplify noise in the image where as other filters like the Sobel serve to reduce noise in images. The Unsharp filter is a simple sharpening operator that enhances the edges and other high frequency components in an image via a process which removes an unsharp/ smoothed side of an image from the original image. The filter relies on the convolution kernel to conduct spatial filtering. The Unsharp Mask were used on 3 various image types, that is the X-ray, bits and altered rose, with various span values of 5, 50 and 110. The results were smoothed and Unsharp Masked images with different appearances obtained. For the smoothed images with span values equal to 5 appeared blurred and the quality of the images reduced compared to those of original images. By increasing the span value to 50 and 110, the blurring becomes more concentrated in the middle of the image. Because of these high span values, the inside and middle details become invincible and the edges of the images become sharper with more details discernible over the edges. When the Unsharp Mask is used on smoothed images with various span values, in this case 5, 50 and 110, the results are Unsharp Masked images. The images with span value of 5 have enhanced and detectable image edges in the 3 images and the backgrounds become grey and the edges much sharper. With images with span value of 50, the mask covered the borders of the images and the images’ center details were clearer and with higher contrast than smoothed images with similar span values which were blurred. The Unsharp Masked images with span values of 110 were much better and with more contrast for the image borders and the edges. They have more blackness to them with equally superior center details. Read More